What is the Chain of Thought Podcast?

Updated

Chain of Thought is a podcast for engineers, founders, and operators building AI, hosted by Conor Bronsdon. It focuses on AI infrastructure, agents, and evaluation, and features guests from companies including NVIDIA, Google DeepMind, and AMD. If you want production-grounded conversations about how AI systems actually get built and run, rather than high-level hype, this show is aimed at you.

What the show covers

The stated focus is "AI infrastructure, agents, and evals," with the broader framing of "what's changing in AI infrastructure." Episodes are organized around topics such as:

  • AI Agents
  • AI Evaluation & Reliability
  • Enterprise AI
  • Open Source AI
  • RAG & Retrieval
  • AI Hardware
  • AI Observability
  • AI Coding
  • Agent Memory
  • Multimodal AI
  • Context Management

Recent episodes illustrate the range:

Episode Topic Guest
EP 77 Switching models without starting over Kimberly Logan, Walrus Protocol
EP 76 AI engineering metrics in the enterprise Andrea Malagodi, Sonar
EP 75 Why context alone isn't enough for enterprise AI agents Kapil Chhabra, WisdomAI
EP 74 Product engineers deciding what to build Laurie Voss, Arize

Who it's for

The show describes its audience as engineers, founders, and operators shipping AI. The content assumes some familiarity with building or operating AI systems — it's a builder-focused show, not an introductory AI primer.

Formats and where to listen

  • Podcast episodes — available via YouTube, Spotify, Apple Podcasts, and RSS.
  • Show notes and transcripts — each episode links to written notes and a transcript.
  • Newsletter — essays drawn from the conversations, plus new episodes, delivered up to once a week.
  • Explainers — a library of plain-answer articles on builder questions, each linked to relevant guests. Examples include "Should you evaluate AI with an LLM-as-a-judge or with human review?" and "Should you build a single agent or a multi-agent system?"

Scale and activity

As of October 8, 2026, the show reports:

  • 77 episodes
  • 164,924+ downloads and full-episode views
  • 3,050 YouTube subscribers
  • 1,200+ newsletter subscribers
  • 17K+ host following on LinkedIn and X

Is it worth following?

It's a reasonable fit if you want practitioner-level discussion of AI infrastructure, agents, and evaluation, with guests from major AI hardware and research organizations. The combination of episodes, transcripts, and written explainers makes it usable both as listening material and as a reference you can search when you hit a specific design question. The newsletter is free to subscribe to and described as up to once a week with unsubscribe anytime.